Reference no: EM133916240
Data-driven Forecasting
Assessment - Fitting and evaluating time series models
Assessment Instructions
Part A Report: Forecasting for Decision-Making: A Data-Driven Approach
Case Scenario:
You are a Data Scientist at a global analytics firm and have been assigned to forecast trends in a unique timeseries of monthly values (each student will receive different data). Your forecasts will help business leaders optimise decision-making in areas such as demand planning, climate predictions, financial projections, or consumer trends.
Your task is to apply a series of forecasting techniques and produce a report explaining your methodology, results, and recommendations.
You will be provided with a dataset on Monday of Week 7.
Your report should be well-structured, professionally formatted, and include clear visualisations to communicate findings effectively.
You can use software of your choice such as Excel, Exploratory, Orange, Python, or Tableau Public to assist.
Exponential Smoothing & Holt-Winters Method
Apply Simple Exponential Smoothing and Holt-Winters (Triple Exponential Smoothing) and forecast for one year ahead. Calculate the MASE for Holt-Winters over the final year, where you define a monthly seasonal naive forecast as the value in the latest corresponding month.
Interpret all of the analysis above.
Prophet and correlogram
Fit Prophet to the time series with additive seasonals and with multiplicative seasonals. Calculate the RMSE. Continue with the option that gives the lower RMSE. Get top-rated assignment help now.
Identify seasonality patterns and trends and discuss their impact on forecasting decisions.
Obtain the Remainder series from your Prophet fit. Calculate the ACF and PACF of the Remainder series. Explain the role of ACF and PACF plots in time series forecasting.
Interpret all of the analysis above.
Seasonal ARIMA Modelling
Implement seasonal ARIMA models.
Discuss the model selection criteria you use to choose a suitable model. State the order of your chosen model together with its RMSE
Forecast for 1 year ahead and provide 68% limits of prediction.
Compare your chosen seasonal ARIMA results to Holt-Winters and state which model you would recommend in this context.
Var models
What is a VAR model, and when is it used in forecasting?
Find a real-world application of VAR and discuss how VAR can be implemented.
Explain the concept of Granger causality and its application in time series forecasting.
Is there evidence of Granger causality in the VAR application you found?
Professionalism
Your report should be clear, well-structured, and formatted professionally to effectively communicate your forecasting insights. Include graphs, tables, and statistical outputs to support your discussion. Ensure clear labelling and professional formatting.
Your referencing must be correct and without the use of Generative AI in the report.